the current python reference solution produces nonsense for those inputs, and those inputs might not make very good sense either.
js produces similar input cases, so it would be nice if author could fix that first before moving on to the other languages

I encountered the same problem. There seems to be an issue with the random tests where the last layers of a landscape have too few characters. The solution for those random tests ignore the coloumns with missing characters alltogether.
Example:
Input string:
"ABC\nDEF\nG"
The provided solution would be:
"A\nD\nG"

Hi, with Python I have trouble with some random test, for instance with this landscape:
c tHu6
Yd7+f
*

my solution fail:
'cAtHu6\n Yd7+f\n*'

kata response should equal:
'c\n \n*'

I don't understand why.
Am I miss something ?

Done + updated to NodeJS 14 and new framework.

random tests should be more complex, they can be solved with a non optimal solution unlike the last two sample tests

elegant solution, but does not work for some cases

for example : "(12x) 456-7x90"

Perhaps it should be made clear that accented characters don't need to be encrypted,
or test encryption with accented characters so that we can understand it.

I had trouble with the French sentences,
because in these cases the encrypted message is not checked,
we just test our ability to decrypt what we have just encrypted.

As these were the last two tests to fail while all the others passed,
I didn't realise that it was my encryption that was wrong.

I finally solved the problem by using a global variable dictionary,
which is not a good method. :)

Translated with www.DeepL.com/Translator (free version)

there is a mistake:
"count = Counter(cars)" should not be in the "for index" loop.

I don't know why, but this solution works well althought it can happend that random test do not pass.

I just completed it in python, I don't think I have done it in a refined way, but it's work fine for me.